Biography

I am currently an MPhil student in AI under the supervision of Prof. Li Liu at HKUST(GZ) and the co-supervision of Prof. Yi R. (May) Fung at HKUST. Previously, after getting my MSc CS degree from HKU, I used to work as a full-time research assistant at Applied Security, Trust And Privacy Lab for Enterprise of PolyU under the supervision of Prof. Haibo Hu. During my undergraduate time at ECNU, I also participated in part-time research at Shanghai Key Laboratory of Trustworthy Computing under the supervision of Prof. Dehui Du. It is also my great pleasure to work with close co-authors Zi Liang, Weilin Lin, and Nanjun Zhou.

Interests
  • AI Security
  • Adversarial/Backdoor Attack
  • Adversarial Training
  • LLM Alignment
Education

  • M.Phil. in Artificial Intelligence, 2024-Now

    The Hong Kong University of Science and Technology (Guangzhou)

  • M.Sc. in Computer Science, 2022-2024

    The University of Hong Kong

  • B.Eng. in Software Engineering, 2018-2022

    East China Normal University

News

[2025.06] Our paper on accuracy-robustness trade-off in adversarial training has been accepted by ICCV'25. Many thanks to Prof. Liu!

[2025.05] Our paper on model extraction attacks for LLMs has been accepted by ACL'25. Congratulations to Zi!

[2024.09] I start my M.Phil. study at HKUST(GZ). Please feel free to say hi if you see me around the campus~

[2024.07] Our paper on adversarial attack against time series classification has been accepted by ECAI'24. Many thanks to Prof. Du, Prof. Hu, Zi, and Yuanhao!

[2024.06] My RA journey at ASTAPLE Lab has come to a close. Thanks to everyone for all the help and support, I had a great time :)

Publications

Yanyun Wang, Li Liu (2025). Failure Cases Are Better Learned But Boundary Says Sorry: Facilitating Smooth Perception Change for Accuracy-Robustness Trade-Off in Adversarial Training. In International Conference on Computer Vision (ICCV'25). Accepted to appear.

Yanyun Wang, Li Liu, Zi Liang, Yi R. (May) Fung, Qingqing Ye, Haibo Hu (2025). New Paradigm of Adversarial Training: Breaking Inherent Trade-Off between Accuracy and Robustness via Dummy Classes. ArXiv Preprint.

PDF Code

Weilin Lin, Nanjun Zhou, Yanyun Wang, Jianze Li, Hui Xiong, Li Liu (2025). BackdoorDM: A Comprehensive Benchmark for Backdoor Learning in Diffusion Model. ArXiv Preprint.

PDF Code

Zi Liang, Qingqing Ye, Yanyun Wang, Sen Zhang, Yaxin Xiao, Ronghua Li, Jianliang Xu, Haibo Hu (2025). "Yes, My LoRD." Guiding Language Model Extraction with Locality Reinforced Distillation. In Annual Meeting of the Association for Computational Linguistics (ACL'25). Main conference.

PDF Code

Yanyun Wang, Dehui Du, Haibo Hu, Zi Liang, Yuanhao Liu (2024). TSFool: Crafting Highly-Imperceptible Adversarial Time Series Through Multi-Objective Attack. In European Conference on Artificial Intelligence (ECAI'24). Oral.

PDF Code Slides

Yanyun Wang, Dehui Du, Yuanhao Liu (2023). Meta Pattern Concern Score: A Novel Evaluation Measure with Human Values for Multi-classifiers. In IEEE International Conference on Systems, Man, and Cybernetics (SMC'23).

PDF Code Slides

Mingjun Ma, Dehui Du, Yuanhao Liu, Yanyun Wang, Yiyang Li (2022). Efficient Adversarial Sequence Generation for RNN with Symbolic Weighted Finite Automata. In SafeAI Workshop @ AAAI Conference on Artificial Intelligence (AAAI'22). Best Paper Award Nomination.

PDF Slides Video

Posts

Reading Note of the Paper "A Survey on Neural Speech Synthesis"
Text-to-speech (TTS), also known as speech synthesis, aims to synthesize intelligible and natural speech from text. Neural TTS adopts DNNs as the model backbone for speech synthesis.
Yanyun Wang
Last updated on Mar 17, 2025 1 min read
Reading Note of the Paper "A Survey on Neural Speech Synthesis"
Reading Note of the Paper "Membership Inference Attacks on Machine Learning - A Survey"
Membership Inference Attack (MIA) is to infer whether a data record was used to train a target model or not. MIAs on ML models can directly lead to a privacy breach.
Yanyun Wang
Last updated on Jun 17, 2023 1 min read
Reading Note of the Paper "Membership Inference Attacks on Machine Learning - A Survey"
Reading Note of the Paper "Survey - Leakage and Privacy at Inference Time"
Leakage of data from publicly available ML models is an area of growing significance since it can draw on multiple sources of data, including users’ sensitive data.
Yanyun Wang
Last updated on Jul 1, 2023 1 min read
Reading Note of the Paper "Survey - Leakage and Privacy at Inference Time"
See all posts

Honors & Awards

Excellent Graduate Award, East China Normal University, 2022

Excellent Bachelor's Degree Thesis Award, East China Normal University, 2022

Excellent Undergraduate Student, East China Normal University, 2020 & 2021

First / Second-class Scholarship, East China Normal University, 2020 / 2021

Services

Conference Reviewer - CVPR (2025, 2024), ICCV (2025), AAAI (2025)

Journal Reviewer - TKDE
“All human wisdom is contained in two words: Wait and Hope.” ― Alexandre Dumas